Ultimate HDR?

[wolfnowl]

Hi Folks:

Came across this image on NASA's photo page.  It's a mix of 28 different exposures taken during the recent solar eclipse: http://apod.nasa.gov/apod/ap080808.html

Mike.

[Tim Gray]

cools shot - wouldn't have guessed that's what it would look like.

[francois]

Thanks for sharing... Looks almost like it's a software generated image.

[Guillermo Luijk]

Interesting and beautiful picture.

However, if it is true that the dynamic range of that particular scene was "brightness range of over 10,000 to 1", that means LOG(10000)/LOG(2)=13,29 f-stops of DR. I cannot see then any need for 28 captures. Just 3 of them, 3 f-stops apart would have sufficed with any of our digital cameras.

Looking at the texture of the moon compared to the solar radiance I think the 10000:1 ratio is quite wrong, and the real contrast was higher.

BR

[BradSmith]

Note that the site says the shading in the corona was 10,000 to 1, not the entire image.   Depending on where you define the corona to be, that might be correct.

Interesting and beautiful picture.

However, if it is true that the dynamic range of that particular scene was "brightness range of over 10,000 to 1", that means LOG(10000)/LOG(2)=13,29 f-stops of DR. I cannot see then any need for 28 captures. Just 3 of them, 3 f-stops apart would have sufficed with any of our digital cameras.

Looking at the texture of the moon compared to the solar radiance I think the 10000:1 ratio is quite wrong, and the real contrast was higher.

BR
[a href=\"index.php?act=findpost&pid=216512\"][{POST_SNAPBACK}][/a]

[Ray]

Interesting and beautiful picture.

However, if it is true that the dynamic range of that particular scene was "brightness range of over 10,000 to 1", that means LOG(10000)/LOG(2)=13,29 f-stops of DR. I cannot see then any need for 28 captures. Just 3 of them, 3 f-stops apart would have sufficed with any of our digital cameras.

Looking at the texture of the moon compared to the solar radiance I think the 10000:1 ratio is quite wrong, and the real contrast was higher.

BR
[a href=\"index.php?act=findpost&pid=216512\"][{POST_SNAPBACK}][/a]

Guillermo,
I wonder if the great number of images has something to do with the limitations of Photoshop's Merge to HDR. I often find when attempting to merge 3 images which are 3-4 stops apart that I get strange discolorations in the sky. This usually occurs when the longest exposure has a severly blown out sky (for example) and the shortest exposure has a correctly exposed sky. In such circumstances, I usually have to discard the longest exposure, which provides the best detail in the shadows, and content myself with a more modest DR increase of 1.5 to 2 stops by merging just the two shorter exposures.

In other words, it's better to merge a larger number of images with a smaller exposure difference. I wish my camera could autobracket 5 exposures rather than just 3.

I believe your noise reduction method has addressed this problem

[Guillermo Luijk]

In other words, it's better to merge a larger number of images with a smaller exposure difference. I wish my camera could autobracket 5 exposures rather than just 3.

I believe your noise reduction method has addressed this problem

A gap of 3-4 f-stops should not be a problem for any camera. Just a simple pixel selection blending would provide a good result image if alignment is good with such gaps.

I did some tests with PS blend for HDR, using the same 3 source TIFF files in PS and in Zero Noise, and while ZN takes all pixels from the dark areas from the most exposed shot (and that means optimum SNR and no danger of shaprness loss due to missalignment), Photoshop takes information of at least two images in the set, providing a worse SNR for noise contamintation from less exposed shots:




I have come to the conclusion that using 3 bracketed shots {0EV, +2EV, +4EV} is my preferred scheme for any situation. In present cameras you cannot bracket more without touching the camera (very important for proper alignment), but you actually don't need it to capture all DR as long as the 0EV shot is properly ETTR'ed. Results are very good even with a modest 350D; any newer camera should be even better.